NFS - i guess we will learn with the roadmap how the plan is here ..?


hmmm - i was thinking about how to best add the NFS-capabilities to pySafe …

but since @oetyng seems to have suggested a way more powerful systematic than the current solution i’m not sure if the entire api will change and if it makes sense to bother implementing this …

if i’m not mistaken the only advantage of using the NFS interface is that there is a file-datatype

 uint64_t size;
 int64_t created_sec;
 uint32_t created_nsec;
 int64_t modified_sec;
 uint32_t modified_nsec;
 uint8_t* user_metadata_ptr;
 uintptr_t user_metadata_len;
 uintptr_t user_metadata_cap;
 XorNameArray data_map_name;
} File;

so it’s basically an immutable data with additional metadata;

a ‘directory’ is just a mutable data and the keys in it probably are the filenames that link to the file objects …?

since i need to manage my directories/mutables then myself anyway (i didn’t see a way to create directories in the api?) i’m not even sure how the plan here was to operate the NFS …?

should i stop thinking because everything will change anyway …? or should i implement the current interface because just the stuff under the hood will change?


I think the principle advantage is interperoperabilty. This need not be NFS, but could be something that replaces it, maybe even based on RDF, but for the moment the only native API standard we have is NFS, which is a strong candidate because it’s a well understood metaphor even if the implementation is currently limited. I’m interested to see if Maidsafe and others propose to change this significantly or just refine it so that it is not so hard to build on it. So I’ve been awaiting discussion and proposals on this too.

The delay is unfortunate because it inhibits development (as we both demonstrate) so I hope we can get clarity on this and Appendable Data ASAP.

1 Like

If directories would be part of the interface I would agree that it makes a lot of sense - without it every dev will be required to come up with a solution for this and therefore I would tend to ignore the whole nfs interface in python and look for a more powerful python specific solution for it…


1 Like

They are, but this has never been documented and so took me a long time to discover (literally years!) See:

I have not taken this as final though, because:

a) without being documented it is not reliable IMO.

b) it is unsatisfactory (again IMO) and therefore should be subject to discussion before being finalised, but we can’t discuss something that isn’t documented.

c) the implementation or how best to use it might be affected by changes introduced for Appendable Data.

Hopefully with the roadmap in place we’ll get an idea of when this might be discussed or finalised.